Rate-Distortion Optimization for JPEG2000

During our Ph.D. studies we studied rate-distortion optimization mechanisms employed by the JPEG2000 standard. Our Ph.D. dissertation introduces three rate control methods to enhance the quality scalability features of JPEG2000. Then, this research line lead to rate-distortion optimized strategies for the layer allocation and truncation of JPEG2000 codestreams. The main ideas behind these works are summarized below.

MODEL-BASED JPEG2000 RATE CONTROL METHODS (Ph.D. thesis)

ABSTRACT: This work is focused on the quality scalability of the JPEG2000 image compression standard. Quality scalability is an important feature that allows the truncation of the codestream at different bit-rates without penalizing the coding performance. Quality scalability is also fundamental in interactive image transmissions to allow the delivery of Windows of Interest (WOI) at increasing qualities.

JPEG2000 achieves quality scalability through the rate control method used in the encoding process, which embeds quality layers to the code-stream. In some scenarios, this architecture might raise two drawbacks: on the one hand, when the coding process finishes, the number and bit-rates of quality layers are fixed, causing a lack of quality scalability to code-streams encoded with a single or few quality layers. On the other hand, the rate control method constructs quality layers considering the rate-distortion optimization of the complete image, and this might not allocate the quality layers adequately for the delivery of a WOI at increasing qualities.

This work introduces three rate control methods that supply quality scalability for WOIs, or for the complete image, even if the code-stream contains a single or few quality layers. The first method is based on a simple Coding Passes Interleaving (CPI) that models the rate-distortion through a classical approach. An accurate analysis of CPI motivates the second rate control method, which introduces simple modifications to CPI based on a Reverse subband scanning Order and coding passes Concatenation (ROC). The third method benefits from the rate-distortion models of CPI and ROC, developing an approach based on a novel Characterization of the Rate--Distortion slope (CoRD) that estimates the rate-distortion of the code-blocks within a subband.

Experimental results suggest that CPI and ROC are able to supply quality scalability to code-streams, even if they contain a single or few quality layers, achieving a coding performance almost equivalent to the one obtained with the use of quality layers. However, the results of CPI are unbalanced among bit-rates, and ROC presents an irregular coding performance for some corpus of images.CoRD outperforms CPI and ROC achieving a well-balanced and regular results and, in addition, it obtains a slightly better coding performance than the one achieved with the use of quality layers. The computational complexity of CPI, ROC and CoRD is negligible in practice, making them suitable to control interactive image transmissions.

Ph.D. DISSERTATION: doc.pdf4.1M
CoRD PAPER: F. Auli-Llinas and J. Serra-Sagrista, JPEG2000 quality scalability without quality layers, IEEE Trans. Circuits Syst. Video Technol., vol. 18, no. 7, pp. 923-936, Jul. 2008. (DOI:10.1109/TCSVT.2008.920748, doc.pdf816K)
ROC PAPER: F. Auli-Llinas and J. Serra-Sagrista, Low complexity JPEG2000 rate control through reverse subband scanning order and coding passes concatenation, IEEE Signal Process. Lett., vol. 14, no. 4, pp. 251-254, Apr. 2007. (DOI:10.1109/LSP.2006.884900, doc.pdf346K)
CPI PAPER: F. Auli-Llinas, J. Serra-Sagrista, J.L. Monteagudo-Pereira, and J. Bartrina-Rapesta, Efficient rate control for JPEG2000 coder and decoder, in Proc. IEEE Data Compression Conference, Mar. 2006, pp. 282-291.
PRESENTATION: slides.pdf7.1M, slides.pptx36M
STRATEGIES FOR JPEG2000 QUALITY LAYER ALLOCATION AND TRUNCATION (further work)

ABSTRACT: Employing the ideas and theory behind the above rate control methods, we developed an allocation strategy of JPEG2000 quality layers based on the hypothesis that the fractional bitplane coder of JPEG2000 already generates optimal truncation points for the overall optimization of the image. Through these overall optimal truncation points, the proposed strategy is able to allocate quality layers without employing rate-distortion optimization techniques, to self-determine the density and distribution of quality layers, and to reduce the computational load of the encoder. Experimental results suggest that the proposed method constructs near-optimal codestreams in the rate-distortion sense, achieving a similar coding performance as compared with the common PCRD-based approach.

Also, there are some applications such as interactive image transmission, video streaming, or transcoding that demand layer fragmentation. The common approach to truncate layers is to keep the initial prefix of the to-be-truncated layer, which may greatly penalize the quality of decoded images. The following paper introduces a Block-Wise Layer Truncation (BWLT) that, requiring negligible computational costs, enhances the quality scalability of compressed JPEG2000 images. The main insight behind BWLT is to dismantle and reassemble the to-be-fragmented layer by selecting the most relevant codestream segments of codeblocks within that layer. The selection process is conceived from a rate-distortion model that finely estimates rate-distortion contributions of codeblocks. Experimental results suggest that BWLT achieves near-optimal performance even when the codestream contains a single quality layer.

LAYERS ALLOCATION PAPER: F. Auli-Llinas, J. Bartrina-Rapesta, and J. Serra-Sagrista, Self-conducted allocation strategy of quality layers for JPEG2000, EURASIP Journal on Advances in Signal Processing, vol. 2008, pp. 1-7, 2008, article ID 728794. (DOI:10.1155/2008/728794, doc.pdf669K)
BWLT PAPER: F. Auli-Llinas, J. Bartrina-Rapesta, and J. Serra-Sagrista, Enhanced JPEG2000 Quality Scalability through Block-Wise Layer Truncation, EURASIP Journal on Advances in Signal Processing, vol. 2010, pp. 1-11, 2010, article ID 803542. (DOI:10.1155/2010/803542, doc.pdf3M)
BWLT PRESENTATION: slides.pptx19M